Calibration of microscopic traffic-flow models using multiple data sources

Area of Science:

• Traffic flow dynamics
• Microscopic traffic modeling
• Statistical estimation techniques

Background:

• Parameter identification in microscopic driving models is challenging due to unobservable parameters and limited statistical methods.
• Existing car-following models often lack robust parameter estimation techniques.
• Direct observation of parameters like reaction time is difficult from typical traffic data.

Purpose of the Study:

• To develop a new approach for identifying parameters of car-following models.
• To enable joint estimation of parameters using diverse data sources, including prior information.
• To facilitate statistical analysis and cross-comparison of different model complexities.

Main Methods:

• Generalization of the maximum-likelihood estimation approach.
• Joint parameter estimation utilizing multiple data sources (e.g., helicopter-collected trajectories, driving simulator data).
• Application of the likelihood-ratio test for comparing models with varying complexity.

Main Results:

• The proposed approach allows for joint estimation of car-following model parameters from heterogeneous data.
• Statistical analysis of parameter estimates, including standard errors and correlations, is enabled.
• The likelihood-ratio test effectively compares models based on parameter count and performance.

Conclusions:

• The developed method offers a robust framework for parameter identification in microscopic driving models.
• The approach improves the accuracy and reliability of parameter estimation by integrating multiple data sources.
• This methodology provides a valuable tool for advancing the understanding and simulation of traffic dynamics.